Minisatellite Variants Generated in Yeast Meiosis Involve DNA Removal During Gene Conversion

Corresponding author: Rhona H. Borts, Department of Genetics, University of Leicester, Leicester LE1 7RH, United Kingdom., rhb7{at}le.ac.uk (E-mail)

Communicating editor: M. LICHTEN

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Two yeast minisatellite alleles were cloned and inserted into a genetically defined interval in Saccharomyces cerevisiae. Analysis of flanking markers in combination with sequencing allowed the determination of the meiotic events that produced minisatellites with altered lengths. Tetrad analysis revealed that gene conversions, deletions, or complex combinations of both were involved in producing minisatellite variants. Similar changes were obtained following selection for nearby gene conversions or crossovers among random spores. The largest class of events involving the minisatellite was a 3:1 segregation of parental-size alleles, a class that would have been missed in all previous studies of minisatellites. Comparison of the sequences of the parental and novel alleles revealed that DNA must have been removed from the recipient array while a newly synthesized copy of donor array sequences was inserted. The length of inserted sequences did not appear to be constrained by the length of DNA that was removed. In cases where one or both sides of the insertion could be determined, the insertion endpoints were consistent with the suggestion that the event was mediated by alignment of homologous stretches of donor/recipient DNA.

VARIABLE number tandem repeats (VNTRs), such as minisatellites and microsatellites, have become valuable tools for genome mapping (NAKAMURA et al. 1987 ), forensic identification (GILL et al. 1985 ; JEFFREYS et al. 1985A , JEFFREYS et al. 1985C ), and population analysis (WETTON et al. 1987 ). VNTRs may vary both in length (number of repeat units) and in repeat unit sequences (JEFFREYS et al. 1985B ). Minisatellites are distinguished from microsatellites by their larger repeat unit, ranging from 8 to 100 bp in length. The repeat units often exhibit sequence polymorphisms that distinguish one unit from the others in the array (WONG et al. 1986 , WONG et al. 1987 ; JEFFREYS et al. 1990 ) such that different minisatellite alleles may consist of a different sequential order of repeat units. The combination of these variables can produce >90% observable heterozygosity (WONG et al. 1986 , WONG et al. 1987 ). The 36 base-pair repeat (36 BPR) of Saccharomyces cerevisiae is a typical subtelomeric minisatellite with alleles that differ in the number, order, and sequence of repeat units (HOROWITZ and HABER 1984 ).

A number of lines of evidence suggest that the high level of allelic minisatellite polymorphism seen in humans is meiotically derived (JEFFREYS et al. 1994 ; JEFFREYS and NEUMANN 1997 ). Rates of change are much higher in germline cells than in somatic tissues such as blood (JEFFREYS et al. 1994 ). The distribution of sizes and the internal structure of the alleles recovered are different in the two cell types, indicating mechanistic differences in how they arise (JEFFREYS et al. 1994 ; JEFFREYS and NEUMANN 1997 ). Finally, the germline events characterized do not appear to be clonally derived (JEFFREYS et al. 1994 ). This observation is more consistent with a meiotic origin rather than an origin during the mitotic divisions of germline development. The existence of cis-acting elements that elevate rates only in germline tissue are also taken as evidence for a meiotic origin of minisatellite size changes (JEFFREYS et al. 1998A ). Proposed mechanisms to explain changes in minisatellite size include unequal crossing over (SMITH 1976 ; JEFFREYS et al. 1998A , JEFFREYS et al. 1998B ), gene conversion (JEFFREYS et al. 1994 ), and replication slippage (LEVINSON and GUTMAN 1987A , LEVINSON and GUTMAN 1987B ; SCHLOTTERER and TAUTZ 1992 ). Distinguishing between such mechanisms is often difficult. Detecting an unequal crossover requires the recovery of both recombinant meiotic products. To detect gene conversion, the nonreciprocal transfer of sequence information, such that alleles do not segregate in a normal 2:2 pattern, requires the analysis of all four meiotic products. In mammalian systems, segregants where the information of one allele is flanked by sequences from the other parental allele have been interpreted to be the result of gene conversion. However, such recombinants may also arise from a double crossover or other more complicated events. Replication slippage during the premeiotic DNA synthesis may also cause changes in sequence arrangement, but without transfer of information between homologues. In most systems it is impossible to distinguish forward slippage during replication from an event involving an intrachromosomal deletion by recombination (for an exception see RAY et al. 1988 ).

The present consensus from human and yeast studies is that the mechanism by which minisatellites change size resembles gene conversion as characterized in the fungi (BUARD and VERGNAUD 1994 ; JEFFREYS et al. 1994 , JEFFREYS et al. 1998B ; APPELGREN et al. 1997 , APPELGREN et al. 1999 ; DEBRAUWERE et al. 1999 ). For example, studies of human minisatellites have noted that alterations in minisatellites occur frequently at one end of an array, with few or no events toward the other end (JEFFREYS et al. 1990 , JEFFREYS et al. 1991 ; ARMOUR et al. 1993 ). This clustering of events is reminiscent of the gradients of gene conversion (polarity) observed at many loci in fungi (reviewed in PAQUES and HABER 1999 ; PETES et al. 1991 ), where the frequency of events decreases with distance from the initiating double-strand break (NICOLAS et al. 1989 ; SUN et al. 1989 ; CAO et al. 1990 ). A putative hotspot that, when present, confers elevated levels of meiotic recombination has been identified upstream of a human minisatellite and has been proposed to be the site of initiation for minisatellite changes (JEFFREYS et al. 1998A ).

Due to the difficulties of characterizing recombination events in humans, model systems where different human minisatellites have been inserted into yeast have been developed (APPELGREN et al. 1997 , APPELGREN et al. 1999 ; DEBRAUWERE et al. 1999 ). These studies find that the types of events recovered are similar to those found in humans and have many of the attributes of meiotic recombination found using conventional alleles in yeast. For example, APPELGREN et al. 1999 unequivocally demonstrate gene conversion and a polarity gradient, while DEBRAUWERE et al. 1999 demonstrate that heteroduplex DNA is formed during minisatellite recombination. They further show that minisatellite recombination is dependent on the production of double-strand breaks and that the frequency of double-strand breaks correlates well with the frequency of recombination.

We present here a model system for studying the generation of meiotic variants in a naturally occurring minisatellite of the yeast, S. cerevisiae. Using a genetically defined chromosomal interval containing the 36 BPR minisatellite, we examined and mapped recombination events that occur both within and flanking these elements. Sequence analysis was used to characterize rearranged minisatellite arrays. By tetrad analysis we identified events resembling gene conversion, deletion (or replication slippage), and a complex combination of both. No reciprocal events were detected, supporting previous findings that unequal crossing over is not the major mechanism of minisatellite change (WOLFF et al. 1988 , WOLFF et al. 1989 ; JEFFREYS et al. 1990 ). APPELGREN et al. 1999 have recently reported similar results in their analysis of the MS32 minisatellite inserted adjacent to the LEU2 hotspot. We obtained additional minisatellite rearrangements by selecting for recombination events in the vicinity of the minisatellite inserts. Analysis of these events indicated that there is an association of changes in minisatellite size with crossing over and gene conversions.

Five genes were mutated to determine their effect on minisatellite recombination events. Four encode components of the mismatch repair system (Msh2p, Msh3p, Mlh1p, and Pms1p), and the fifth encodes the RecQ homologue, Sgs1p (WATT et al. 1995 , WATT et al. 1996 ; LU et al. 1996 ). Mismatch repair proteins were chosen because they are involved in the repair of heteroduplex DNA during meiotic recombination (reviewed in BORTS et al. 2000 ; CROUSE 1998 ) and in recombination between diverged sequences (BORTS and HABER 1987 ; BORTS et al. 1990 ; CHAMBERS et al. 1996 ; HUNTER et al. 1996 ; CHEN and JINKS-ROBERTSON 1999 ). No changes in the frequency of minisatellite alteration were found in any of the mismatch repair mutants analyzed. This is in contrast to the significant destabilization found for a CEB1 minisatellite in a mismatch repair defective strain (DEBRAUWERE et al. 1999 ).

The effect of mutation in SGS1 was examined because two of the three known human RecQ homologues, BLM and WRN, are required for genomic stability (ROSIN and GERMAN 1985 ; GEBHART et al. 1988 ). Interestingly, there have been conflicting reports on the stability (FOUCAULT et al. 1996 ) or instability (GRODEN and GERMAN 1992 ) of minisatellites in cells from Bloom's syndrome patients. No change in stability was observed in this mutant background.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Strains:
All constructs were made in the following two derivatives of the yeast strain Y55 (MCCUSKER and HABER 1998): Y55.2172 = HO MATa lys2-d leu2 ura3-n ade1-1 his6-1 can1 and Y55.2203 = HO MAT leu2 ura3-n cyh2 met13-4.

Genetic procedures:
Yeast manipulations and media were as described (ROSE et al. 1990 ). Strains were grown at 30° on YEPD or synthetic complete media lacking one or more nutrient. Sporulation was performed at room temperature as described (HUNTER et al. 1996 ).

PCR amplification:
All PCR amplifications followed the same basic protocol. PCR amplification was performed in NH₄ buffer (supplied by Bioline Ltd., London) with 3.5 mM MgCl₂, 0.15 pMol of each primer, and 0.01 U/µl of Taq polymerase (Bioline). The reaction was heated to 95° for 3 min, then cycled 30 times at 61° for 30 sec, 72° for 1.5 min, and 94° for 20 sec, with a final annealing of 2 min and extension at 72° for 2.5 min. The primer pair prA-F and prA-R have optimal annealing temperatures of 58.5°. For amplification with the primer pair prC-F and prC-R a longer extension time of 2 min during the cycling phase was used.

Plasmid constructs:
Two plasmids, pRED40 and pRED55, were created based on pMJ13 (LICHTEN et al. 1987 ). pMJ13 contains MATa, URA3, and LEU2 sequences. pRED40 and pRED55 were derived from this plasmid by mutation of LEU2 at the Asp718 site (leu2-a) and of URA3 at the NcoI site(ura3-n) as described (BORTS and HABER 1989 ).

The pAB series of plasmids was derived from either pRED40 or pRED55. Two 36 BPR alleles, 12 x 36 BPR and 15 x 36 BPR, were amplified by PCR from the Y' clones pEL-31 (LOUIS and HABER 1992 ) and p69-3' (similar to the previously described HinfI 1-6 clone; HOROWITZ and HABER 1984 ), respectively, using the primers prA-F (5'-GGC GGA AGA TCT TCT GCA GCA GTG ATG AGG ACA GC-3') and prA-R (5'-GCG GGA AGA TCT GAA TTC ATC TTT ATT GGC GTC CTC C-3'). These primers amplified sequences corresponding to 43 bp 5' prior to the 36 BPR array to 23 bp 3' after the array within the subtelomeric Y' sequence. PCR products were digested with BglII (underlined) and purified using Stratagene (La Jolla, CA) PrimeErase Quick Columns. The 12mer derivative was ligated into BamHI-digested pRED40 and pRED55 to create pAB3 and pAB7, respectively. Similarly, the 15mer derivative was used to produce pAB1 and pAB5. The orientation of the 36 BPR array within each of the plasmid constructs was determined by digestion with either PstI or EcoRI, which cleave prA-F or prA-R, respectively (double underlined). The orientation of the 36 BPR was chosen such that after integration into the yeast genome the arrays were oriented in the same direction as those in the native Y' elements. The BamHI site of pBR322 has been shown previously to be very close to a site of double-strand breaks in constructs similar to the ones used here (WU and LICHTEN 1995 ).

Yeast strain construction:
MATa-URA3-LEU2-MATa and MAT-URA3-LEU2-MAT nontandem duplications were created by transformation (BORTS et al. 1986 ; GEITZ et al. 1992 ) into Y55.2172 and Y55.2203. The general structure of the nontandem duplicated MAT locus is illustrated in Fig 1 and the relevant genotypes of specific diploids are listed in Table 1.

View larger version (14K): In this window In a new window Download PPT slide   Figure 1. The MAT-MAT nontandem duplication on S. cerevisiae chromosome III. pBR322 sequences are indicated by the thin line. The 36 BPR array (12mer, ; 15mer, ) was inserted at the BamHI site (B). The distances (in base pairs) between the minisatellite insertion site and the nearest genetic markers are indicated. Restriction sites (HindIII, H; EcoRI, R; SalI, S; and XhoI, X) used in plasmid construction are indicated. The specific configuration of mating-type, minisatellite, and auxotrophic markers for each diploid used is given in Table 1.

Y55.2172 and Y55.2203 derivatives containing mutations in MSH2, MSH3, MLH1, PMS1, or SGS1 were created using the plasmids pII-2::7-7 (REENAN and KOLODNER 1992 ), pGEM7-Zf(+)/CX URA3 (HUNTER and BORTS 1997 ), pmlh1::URA3 (obtained from M. Liskay), pWBK4 (KRAMER et al. 1989B ; HUNTER and BORTS 1997 ), and pPWsgs1 (WATT et al. 1995 ), respectively. These constructs utilize either URA3 or LEU2, which are the markers used to select random spore events, to disrupt the relevant gene. It was thus necessary to further disrupt the URA3 or LEU2 marker with LYS2. Plasmids containing LYS2 disruptions of URA3 or LEU2, pRED223, and pRHB142, respectively, were created and used for this purpose. The mutant strains were then crossed with the minisatellite constructs to obtain the appropriate segregants.

Tetrad analysis:
Spore colonies from dissected tetrads were grown for 3–5 days at 30°. Tetrads containing four viable spores were examined. Genetic analysis was conducted by replica plating onto various synthetic complete media lacking one nutrient. Mating phenotypes were determined by the use of tester strains of known mating type.

Random spore analysis:
Random spores were prepared as described (LICHTEN et al. 1987 ) and grown for 3–6 days at 30° on synthetic complete medium lacking arginine and containing cycloheximide (10 mg/liter) and canavanine (40 mg/liter). Ura⁺ Leu⁺ recombinant spores were selected on a similar medium that also lacked uracil and leucine. Individual clones were tested for mating phenotype and examined by PCR to determine the size of the 36 BPR array. A nonmating phenotype is indicative of a reciprocal crossover, an unequal crossover, coconversion of URA3 or LEU2 and a flanking MAT locus, or meiosis I nondisjunction of chromosome III. Nonmating clones that contained nonparental-size 36 BPR arrays were also examined by Southern analysis (SOUTHERN 1975 ) to determine whether a reciprocal or unequal crossover event resulted in the phenotype. Disomy for chromosome III was tested for by contour-clamped homogeneous electric field analysis (HUNTER et al. 1996 ).

Statistics:
Recombination events in strains with and without the 36 BPR inserts (Table 1 and Fig 2) were compared using a contingency G-test (SOKAL and ROHLF 1969 ). The frequencies of size changes in mutant strains were compared to that in wild type using a standard normal z-test. A goodness-of-fit G-test was used to compare the expected frequency of changes in array size in selected Ura⁺ Leu⁺ recombinants to the observed frequency as described in the text.

View larger version (32K): In this window In a new window Download PPT slide   Figure 2. Recombination events in the MAT-MAT interval. The major classes of genetically detectable recombination events are illustrated. The symbols used are the same as in Fig 1. The rate of occurrence of each class is given for the indicated constructs. * For simplicity, only conversion to wild type for either Ura- or Leu- is illustrated, although all possibilities were recovered. The class of "other" includes unequal sister chromatid exchanges, intrachromosomal deletions, conversions of MAT, and unequal interchromosomal crossovers.

Southern analysis:
All yeast constructs were confirmed by an appropriate restriction endonuclease digestion of genomic DNA (BORTS et al. 1986 ) and Southern analysis (SOUTHERN 1975 ; SAMBROOK et al. 1989 ).

Screening for novel sizes:
To determine the length of the 36 BPR array, colony PCR amplification was performed (PORTER et al. 1993 ) with the following modifications. PCR amplification was conducted on freshly grown yeast cells lifted from a YEPD solid medium and resuspended in 5 µl of Millipore-Q (Millipore, Bedford, MA) water. The cell suspension was heated to 95° for 3 min, and then 15 µl of PCR reaction mix was added. The primers used were prB-F (5'-GGC GAC CAC ACC CGT CC-3'; 54 bp 5' to the array) and prB-R (5'-GAT GCG TCC GGC GTA GAG G-3'; 81 bp 3' to the array). A total of 6 µl of each reaction was analyzed by agarose gel electrophoresis.

Sequencing:
DNA was amplified using a biotinylated primer (indicated by an asterisk). Primer sets of prC*-F (5'-CGG GAT ATC GTC CAT TCC G-3') and prC-R (5'-AGG AAT GGT GCA TGC AAG GA-3') or prC-F and prC*-R were used in 50-µl reactions containing 0.5 µl of genomic DNA (0.1 µg/µl). prC-F and prC-R anneal 225 bp 5' to the array and 237 bp 3' to the array, respectively. The PCR product was isolated using magnetic beads (Dynal, A.S., Oslo). The primer prB-F or prB-R was annealed to the single-strand template containing prC*-R or -F, respectively. Sequencing was then performed with Sequenase (Amersham Pharmacia Biotech, Little Chalfont, United Kingdom) as recommended, including modifications suggested by Dynal, A.S.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The 12mer and 15mer minisatellite arrays:
The 36 BPR (HOROWITZ and HABER 1984 ) of the Y' subtelomeric element (HOROWITZ and HABER 1984 ; LOUIS and HABER 1990 , LOUIS and HABER 1992 ) in S. cerevisiae is a polymorphic minisatellite. Two alleles of the 36 BPR array were cloned and inserted into a genetically defined artificial interval at the MAT locus on chromosome III of S. cerevisiae (BORTS et al. 1986) (Fig 1). The alleles were shown to consist of either 12 (12mer) or 15 (15mer) repeat units of the 36 BPR array by sequence analysis. Each 36 BPR unit is composed of three 12-bp segment types, A, B, and C, the order of which is invariant (HOROWITZ and HABER 1984 ) (Fig 3A). Sequence polymorphisms between each segment type allows a subclassification of the sequences (i.e., variants A1, A2, B1, B2, C1, C2, etc.; Table 2). The first 8 of the 36 BPR units (1A1B1C1–8A3B4C4) and the next A segment (9A3), 304 bp total, are identical between the alleles (see Fig 3A). Within this 304 bp is a tandem duplication, C1-A2B2B2-A3B2C3-A4B3 (underlined in Fig 3A). After 9A3, the 12mer and the 15mer alleles display different repeat organization. A single nucleotide polymorphism (a G in the 12mer and an A in the 15mer) 22 bp upstream of the repeat array allow the centromere proximal ends of the two alleles to be distinguished.

View larger version (28K): In this window In a new window Download PPT slide   Figure 3. Structure of parental and rearranged arrays. (A) A schematic representation of the 12mer and 15mer 36 BPR array sequences by segment types (see Table 1). The regions of identity between the two parental alleles (1A1 to 9A3) are represented by a hatched line. The two 108-bp internal tandem repeats are underlined. A nucleotide polymorphism (G/A) 22 bp proximal to the array is indicated. (B) The sequences of the five rearranged arrays obtained from tetrad analysis. An arrangement of parental arrays that could have produced each novel array is suggested with the recipient DNA as the lower strand in each case. A suggested deletion () of recipient 12mer sequences is shown for NT5. It can be seen that the proposed alignments of the parental arrays overlap with each other in regions where they are homologous. These overlapping regions therefore represent regions of ambiguous origin. For NT4 and NT5 the overlapping regions of the parental arrays illustrated include base discrepancies (white X in black box) between a parental array and the recovered novel array.

Meiotic recombination within the construct:
Analysis of Ura, Leu, and mating phenotypes identified four major classes of recombination events (Fig 2). These include reciprocal crossing over between URA3 and LEU2, reciprocal crossovers between either URA3 or LEU2, and a MAT locus and gene conversion of URA3 or LEU2 with or without a crossover (BORTS et al. 1984 ). Insertion of the 36 BPR does not significantly alter the distribution of these recombination events (Fig 2).

PCR analysis of the 36 BPR of all segregants from the heterozygous diploids identified 36 recombination events involving the 36 BPR (Fig 2). The majority (86%) were 3:1 segregations of parental-size alleles. The remaining five were changes in the number of repeats in the array. Seven of the tetrads with a simple 3:1 gene conversion of the minisatellite were picked at random for further analysis. Sequencing was used to identify the nucleotide at the A/G polymorphism 22 bp upstream of the array and in the last A-segment of the array that distinguished the parental arrays. Three of the seven (one associated with a crossover, one with a gene conversion of LEU2, and one with no other associated event) consisted of the parental configuration of A/G polymorphism and array. The four remaining events (one associated with a crossover, two with a gene conversion of LEU2, and one with no other associated event) had the A/G polymorphism of one parent associated with the array size and sequences of the other. This indicates that the conversion event terminated between the upstream polymorphism and the allele-specific portion of the 36 BPR array. The arrays associated with the nonparental A/G polymorphism were sequenced and showed no sequence alteration, which confirms that conversion encompassed the entire allele. Examination of the sequences from other progeny in each tetrad gave no evidence for any reciprocal sequence changes. These results suggest that additional events, such as gene conversions leading to base changes that do not alter the size of the minisatellite insert, have gone undetected.

Novel minisatellites recovered by tetrad dissection:
The rate of occurrence of size changes of the 36 BPR was 0.5% of meioses (5/1082 tetrads from heterozygous diploids and 1/205 from homozygous diploids). In all six cases the alteration was nonreciprocal.

The sequences of the five novel arrays from the heterozygous diploids are shown schematically in Fig 3 and are described below. In general, parental sequences can be aligned from unique sequences outside of the novel array until mismatches are reached. A summary of the parental configurations that can produce the novel arrays is shown in Table 3. The origins of the first few repeat units of each novel array could not be unambiguously determined because the first eight repeat units and the subsequent A segment are identical between the two parental alleles. However, the origins of the distal sequences could be determined, which, together with the inclusion of the upstream sequence polymorphism, allow us to say that novel tetrads 1, 2, and 5 (NT1, NT2, and NT5) are clearly interchromosomal events while NT3 and -4 could have been derived from intrachromosomal or sister chromatid interactions.

Novel tetrads 1, 2, and 3 are relatively simple events. The NT1 18mer was derived from a 15mer by gene conversion and the insertion of three repeat units resulting in a triplication of the 108-bp tandem repeat. The upstream polymorphism exhibited a 3G:1A segregation pattern indicative of a conversion event while the sequences from 12B3 to the end of the array representing the unique part of the 15mer sequence segregate 2:2. NT3 is similar to NT1 with the exception that the conversion did not encompass the upstream polymorphism. NT3 could also arise as a slippage of 108 bp during replication rather than conversion. The 11mer NT2 can be derived by insertion of 15mer sequences into a 12mer. Sequences from 8A2 to the end of the array are unambiguously of the 15mer (cf. Fig 3A, 12A2 to 15B3) and arise by a transfer of information from a 15mer into a 12mer accompanied by the loss of one repeat unit.

NT4 may have been the result of a simple deletion event mediated by 20 bp of homology either side of the deletion, a gene conversion, or replication slippage during the premeiotic DNA replication. It is not possible to determine how this novel array arose. Of note though, it is possible to extend the parental sequence alignment beyond the first mismatch on one side by 65 bp until subsequent mismatches are encountered (see Fig 3B). Assuming that such an extended homology existed, mismatch repair could remove the single mismatch. Further alignment of the parental sequences in the other events did not reveal any extended sequence homology. NT5 is a complex event and was probably derived by a combination of insertion of 15mer sequences into 12mer sequences accompanied by deletion and mismatch correction. A detailed description of the event can be found in Fig 4.

View larger version (17K): In this window In a new window Download PPT slide   Figure 4. A possible alignment of recipient 12mer and donor 15mer sequences that might produce the observed NT5 sequences are illustrated schematically. Base pair differences between the aligned parental sequences and novel array sequences have been noted (white 1 in black box represents one mismatch within a 12-bp segment). An arrangement of the 12mer sequence where 72 bp are looped out is illustrated. The site of the loop out may be up to 4 bp proximal or 6 bp distal from the position given in the diagram (detailed in the inset with the 10-bp direct repeat underlined). The resultant deletion in 12mer sequences gives no further incorrect nucleotides in comparison with NT5 until 8B2 position 5. The 15mer alignment still results in two incorrect bases in comparison with NT5, but these are obtainable by repair from the 12mer alignment.

Novel minisatellites recovered by random spore analysis:
To determine if there is an association between the generation of minisatellite rearrangements and other recombination events, 576 Ura⁺ Leu⁺ meiotic progeny were selected at random from each of the two heterozygous diploids. Such recombinants are generated by a crossover between URA3 and LEU2, a gene conversion of either ura3-n or leu2-a (with or without an associated crossover), or an unequal crossover between the MAT loci (see Fig 2). The array size was determined for each spore analyzed. The mating phenotype was used to distinguish simple gene conversions of either URA3 or LEU2 from those associated with crossing over, as the latter would yield a haploid nonmating spore. Southern analysis was used to distinguish unequal crossovers between the mating-type loci from reciprocal crossovers. It should be noted that a gene conversion of URA3 or LEU2 that includes one of the flanking mating-type loci could not be distinguished from a crossover between URA3 and LEU2. However, the frequency of these long conversions has been shown to be very low (BORTS and HABER 1987 , BORTS and HABER 1989 ).

From the 1152 Ura⁺ Leu⁺ haploid progeny examined, 21 were identified that had a novel, nonparental-size 36 BPR array. Two events involved an unequal crossover between MAT loci and were excluded from further analysis. Of the 19 (1.6%) remaining Ura⁺ Leu⁺ spores containing novel size arrays all but three (NR7, NR10, NR13) were recombinant for the flanking mating-type loci. The sequences of the arrays are comparable to those observed in the tetrad analysis and are illustrated schematically in Fig 5. Table 3 summarizes the characterization of the simple novel arrays.

View larger version (38K): In this window In a new window Download PPT slide   Figure 5. Eighteen novel array sequences from random spores are represented as in Fig 2. NR1–11 are derived from dAB3 and NR12–18 from dAB4. NR6 is exactly as NR1–5 with the exception that the associated upstream polymorphism is "G." NR7 is exactly like NR1–5 with the exception that the associated upstream polymorphism is a G and that it has a MAT phenotype. NR13 is as NR12 except that the associated upstream polymorphism is an "A" and that it has a MATa phenotype. As with Fig 3, possible alignments of the parental arrays that could give rise to the novel arrangements are indicated. A suggested deletion () is shown for NR16 and base discrepancies between the parental arrays and the novel array are indicated by (white X in black box). The majority of the recovered novel arrays have sequences at their ends consistent with a single event leading to the selected Leu+ Ura+ spore. In other words, they start with the 15mer upstream polymorphism and end with the 12mer array sequence for dAB3 and vice versa for dAB4. Within these are complex insertions of one array's sequences into the other accompanied by deletion of sequences of the recipient array as discussed.

Eleven of the 12 novel size arrays from dAB3 were sequenced. This diploid contains MATa-(ura3-n)-(G-12mer)-(LEU2)-MATa and MAT-(URA3)-(A-15mer)-(leu2-a)-MAT (Table 1). The simple expectation for this configuration of markers is that a novel array would consist of an "A" characteristic of the 15mer parent and the sequences from the distal portion of a 12mer. Eight such recombinants were observed. The three remaining events had the upstream polymorphism and the unique distal sequences of a 12mer (see NR6, 7, and 9, Fig 5 and Table 3). The seven novel arrays derived from dAB4 were also sequenced. dAB4 consists of the arrangement of close flanking markers opposite to that of dAB3 (see Table 1). Of the seven events, four begin with 12mer sequences and end with 15mer sequences (Fig 5 and Table 3), as expected. Two, however, began and ended with 15mer sequences and one with 12mer sequences. Of these three events, NR17 and NR18 (Fig 5) were clearly interallelic as they contained allele-specific sequences from both arrays. These recombinants are therefore of a complex, patchy nature. In total, 14 of the 18 analyzed arrays are unambiguously derived from an interchromosomal minisatellite interaction.

Association between minisatellite rearrangements and recombination of flanking markers:
To determine if minisatellite size changes were associated with other recombination events, we measured their frequency in spores selected to be recombinant for the flanking URA3 and LEU2 genes. If there were no association between an altered minisatellite and other recombination events, it would be expected that the frequency of size changes would remain the same, independent of selection for other recombination events. From the tetrad analysis, we determined the frequency of size changes to be 0.12% (5 minisatellite changes in 4328 spores derived from the heterozygous diploids). Among spores selected to be Ura⁺ Leu⁺ recombinants the frequency was 1.6% (19/1152). This 13-fold enrichment of minisatellite changes demonstrates that there is a significant association between the changes in minisatellite array size and recombination events that produce a Ura⁺ Leu⁺ phenotype (G = 37.3; P < 0.001%).

Novel minisatellites recovered by random spore analysis from mutant strains:
Random spore analysis was carried out in five mutant strains. 576 Ura⁺ Leu⁺ spores from msh2 and 1152 from each of msh3, mlh1, pms1, and sgs1 strains were analyzed (Table 4). No significant difference in the rate of array size change was observed for any mutant background examined. Given the number of spores analyzed, a twofold increase in rate relative to wild type would have been statistically significant.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Minisatellites can be highly polymorphic genetic elements. The mechanisms by which the observed allelic differences are generated are beginning to be defined. The present consensus is that minisatellites alter by a gene conversion mechanism (JEFFREYS et al. 1994 ), which may or may not be accompanied by recombination of flanking markers (JEFFREYS et al. 1998A , JEFFREYS et al. 1998B ). Due to the difficulty of clearly establishing the nature of the recombination event in human systems, yeast has been exploited as a model organism to study these issues. A number of studies have recently been performed that take advantage of the ability to analyze all four meiotic products and that utilize the extensive range of recombination mutants to better characterize the mechanism by which minisatellites change size (APPELGREN et al. 1999 ; DEBRAUWERE et al. 1999 ; this study). In one study, size changes of the human MS32 minisatellite were identified among tetrads whose four meiotic products had been pooled (APPELGREN et al. 1999 ). When a change was identified, all four meiotic products were subsequently analyzed. In another study, changes in size of the human minisatellite CEB1 were shown to be dependent on the formation of a meiotic double-strand break, providing strong evidence for a recombinational mechanism leading to changes in size (DEBRAUWERE et al. 1999 ). A heteroduplex intermediate was also demonstrated and mismatch repair proteins were shown to affect the rate of size changes.

In the study reported here, we analyze all of the meiotic products individually and fully sequence the size changes. This analysis indicates that minisatellites undergo a high rate of simple gene conversion, a type of event that would not have been detected in all previous studies of minisatellite recombination. We also detect the complex events found in other studies and demonstrate that they can be associated with crossing over. We failed to detect an effect of mutation in any of the mismatch repair genes tested, in contrast to that seen in the study of CEB1 (DEBRAUWERE et al. 1999 ).

One observation unique to this study is that six times as many parental-size arrays segregated 3:1 than exhibited a change in size. Half of these 3:1 events were associated with a crossover. Some events involved gene conversions that encompassed the entire array and upstream G/A polymorphism. Other events appear to encompass only one end of the array or the other. The preliminary examination presented here indicates that a simple conversion, as well as changes in size, may be associated with crossovers and/or extended gene conversion tracts. Jeffreys and colleagues have also recently reported exchange of flanking markers without a change in array size (JEFFREYS et al. 1998A , JEFFREYS et al. 1998B ). Although infrequent, such events may be similar to those reported here.

A second conclusion that can be drawn from the tetrad analysis is that substantial amounts of DNA are lost from the recipient array concomitant with insertion of the donor sequences. In the five events analyzed in detail, the minimal amount of sequence lost (calculated from the distance between the nearest sequences of unambiguous parental origin) ranged from 44 to 149 bp. This loss of sequence is difficult to account for by the repair of mismatched heteroduplex involving these sequences.

An intriguing difference between the study reported here and that of the human minisatellites is that the events identified in the 36 BPR were predominantly interallelic. Three out of five minisatellite rearrangements detected in tetrads and 14 out of 18 of the random spore size changes clearly resulted from interallelic events. In contrast, studies of the two different human minisatellites inserted into yeast show a bias toward intraallelic events (APPELGREN et al. 1997 , APPELGREN et al. 1999 ; DEBRAUWERE et al. 1999 ). Such a bias might be indicative of heteroduplex rejection or mismatch-repair-stimulated secondary double-strand breaks of mismatched interhomologue events, both of might then preferentially use intrachromosomal recombination to repair the break. When a mechanism such as single-strand annealing is used to repair the break it can lead to a detectable size change. However, repair from the perfectly homologous sister may not lead to a detectable change. The observation that in the absence of mismatch repair DEBRAUWERE et al. 1999 recover twofold more events, half of which contain unrepaired heteroduplex, is consistent with this proposal.

Another difference among the studies is the variation in rate of size changes detected, which ranged from 0.12% (36 BPR) to 15.6% (MS32) of spores assayed. These differences may reflect the genomic location of the insertion (GOLDMAN and LICHTEN 1996 ; DEBRAUWERE et al. 1999 ) although other explanations are possible. For example, the 20- to 100-fold difference between MS32 and the other minisatellites may reflect the fact that the MS32 rate was measured in a RAD5 defective background (FAN et al. 1996 ; APPELGREN et al. 1997 ). RAD5 encodes a helicase involved in microsatellite instability (JOHNSON et al. 1992 ) and potentially involved in the mechanism of minisatellite changes.

We found no effect on the rate of minisatellite repeat size changes in any of the mismatch repair (pms1, msh2, msh3, mlh1) or helicase (sgs1; WATT et al. 1995 , WATT et al. 1996 ; LU et al. 1996 ) mutants studied. Interestingly, it has recently been shown that the absence of Sgs1p stabilizes an unstable microsatellite (fragile X) inserted into yeast (WHITE et al. 1999 ) during vegetative growth. Although our study is on meiotic minisatellite instability, the lack of an effect of the sgs1 mutation on the 36 BPR supports the notion that the instability of microsatellites and minisatellites may be controlled by different genes (SIA et al. 1997 ; KOKOSKA et al. 1998 ).

The absence of an effect of mismatch repair mutants on the rate of change of the 36 BPR was unexpected. The sequences present in the minisatellite alleles will generate highly mismatched heteroduplex DNA. Because highly mismatched DNA has been shown to be subject to processes that reduce the frequency of interchromosomal recombination (heteroduplex rejection, mismatch-repair-stimulated killing) the prediction is that recombination frequencies would be elevated in the absence of mismatch repair (DOUTRIAUX et al. 1986 ; BORTS and HABER 1987 ; RAYSSIGUIER et al. 1989 ; BORTS et al. 1990 , BORTS et al. 2000 ; CHAMBERS et al. 1996 ; HUNTER et al. 1996 ). Such an effect was seen by DEBRAUWERE et al. 1999 in pms1msh2 double-mutant strains containing CEB1. They also recovered postmeiotic segregations of minisatellite sequences, the hallmark of unrepaired heteroduplex. The increase in rate found in the pms1 msh2 CEB1 strain is consistent with either the recovery of previously undetected events resulting from sister chromatid repair of rejected events or with the recovery of events destroyed by mismatch-repair-stimulated killing. In addition, data suggestive of a role for mismatch-repair- induced recombination in minisatellite size changes can be found in the type of events observed in the MS32 tetrads (APPELGREN et al. 1999 ). Many of the events reported there appear to have arisen from multiple rounds of recombination. Such events are consistent with a model in which the repair of multiple mismatches induces a second recombinational interaction (BORTS and HABER 1987 ; BORTS et al. 1990 ).

There are a number of possible explanations for why DEBRAUWERE et al. 1999 found an effect of mismatch repair genes and none was found in this study. First, DEBRAUWERE et al. 1999 deleted both PMS1 and MSH2 while the genes were deleted individually in this study. It has been suggested that Pms1p and Msh2p may act at least partially independently to impair recombination in the presence of sequence heterozygosity (CHAMBERS et al. 1996 ) and thus it might be necessary to eliminate both genes before an effect would be detectable. Another possible explanation might be that secondary structures formed by the 36 BPR minisatellite during the recombination process might be insensitive to mismatch repair proteins. For example, there is a 15-bp imperfect inverted repeat contained within the 36 BPR unit (HOROWITZ and HABER 1984 ). Palindromic sequences have long been known to escape mismatch repair (NAG et al. 1989 ; NAG and KURST 1997 ) presumably because they form hairpin structures that are not recognized by the mismatch repair system. In addition, single-strand loops of the size postulated to account for many of the events (Fig 6F) are known to be partially (KIRKPATRICK and PETES 1997 ) or completely (WILLIAMSON et al. 1985 ; KRAMER et al. 1989A ) insensitive to the mismatch repair system. It is an intriguing possibility that mechanisms of size change may be sequence or structure dependent and thus may vary from minisatellite to minisatellite.

View larger version (16K): In this window In a new window Download PPT slide   Figure 6. Generation of minisatellite rearrangements by double-strand break repair. After pre-meiotic DNA replication one of the four chromatids suffers a double-strand break and the ends are processed by a 5' to 3' exonuclease that generates two single-strand 3' DNA tails (A). One 3' single-strand end can invade homologous DNA (B) and serve as a primer for DNA synthesis. In the double-strand break repair model (SZOSTAK et al. 1983 ), synthesis displaces one strand of the invaded DNA duplex, making it available to the other 3' single-strand end (C). Once the displaced strand "captures" the other 3' single-strand end by annealing, it could prime DNA synthesis (D) and thus generate a double Holliday junction (E). In the synthesis-dependent single-strand annealing model (HASTINGS 1988 ) following invasion (B) and synthesis, the newly synthesized DNA disassociates from the invaded chromatid and anneals with the other 3' single-strand end (C'). In both models insertion of nonparental sequences into the recipient can occur by misalignment of repeat units between the displaced single-strand DNA and the 3' single-strand end. This can result in unpaired "tails" that must be excised (arrowhead, D, D') before synthesis can repair the break (E, E') and leads to loss of sequence from the recipient. Simple single-strand annealing between repeats (B') can occur following initiation and exposure of the 3' single-strand ends (A). This necessarily results in an intrachromosomal deletion (B'). Misaligned annealing can also generate looped structures such as illustrated in F. Cleavage (arrowhead) of the loop leads to deletion of sequences such as are recovered in NT5.

The results presented above and those from the other minisatellites studied, contribute to our understanding of the mechanisms by which minisatellites change size. The clear interhomologue derivation we see for the majority of size changes is incompatible with replication slippage during the premeiotic DNA synthesis. This, combined with the observation that size changes are dependent on Spo11p-mediated double-strand breaks (DEBRAUWERE et al. 1999 ), lead us to favor recombination as the mechanism by which minisatellites change size. The absence of an effect of mutation in mismatch repair genes leads us to propose a model in which the formation of heteroduplex DNA occurs only within sequences of perfect or nearly perfect homology. Fig 6 illustrates a version of the canonical double-strand break model (SZOSTAK et al. 1983 ) as well as a version of synthesis-dependent single-strand annealing (HASTINGS 1988 ; reviewed in PAQUES and HABER 1999 ). We propose, as do DEBRAUWERE et al. 1999 , that both mechanisms contribute to changes in size of minisatellites. In both models recombination is initiated by a double-strand break in what will be the recipient DNA molecule. This is followed by the removal of one strand by a 5' to 3' exonuclease to generate 3' single-strand tails (SUN et al. 1991 ; reviewed in PAQUES and HABER 1999 ). In a repeated structure, an intrachromosomal deletion can occur by single-strand annealing as is often observed in mitosis (GALLI and SCHIESTL 1995 ) (Fig 6B'). This type of event economically explains the intrachromosomal deletional nature of many minisatellite rearrangements. Noncrossover interhomologue events may also arise by single-strand annealing of newly synthesized DNA. In Fig 6 (C'–E'), a model in which the newly synthesized DNA is processively unwound and anneals with the other 3' single-strand tail is illustrated. Such a mechanism can easily account for array expansions by insertion of sequences from one homologue into the other. Finally, a conventional double Holliday junction structure (COLLINS and NEWLON 1994 ; SCHWACHA and KLECKNER 1994 , SCHWACHA and KLECKNER 1995 ) easily accounts for the 31 gene conversions where parental-size minisatellite arrays segregated 3:1 with or without an associated crossover (Fig 6E).

One observation made in this study not readily accounted for by the conventional models is the net loss of repeats in the recipient arrays (Table 3) detected in the tetrad analysis. This loss of sequence is difficult to explain by simple heteroduplex formation and repair of single base mismatches. However, the loss of recipient DNA could occur at a number of points in the recombination process. For example, sequences could be removed by double-strand exonuclease activity at the time of the breakage by a 3' exonuclease activity acting on the noninvading strand. Another possibility is the degradation of the invading 3' end either prior to or in conjunction with heteroduplex formation. These latter mechanisms all result in gap formation (SZOSTAK et al. 1983 ). Although there has been no physical evidence that either 3' tail is degraded using homologous substrates (LIU et al. 1995 ), no experiments have ever been performed where the interacting sequences are diverged and might provoke such degradation. Loss of DNA in the recipient may also come from cleavage of unpaired 3' single-strand tails (Fig 6D and Fig D'). Such a reaction predicts a role for the Rad1p/Rad10p/Msh2p/Msh3p complex in removing the nonhomologous 3' tail (FISHMAN-LOBELL and HABER 1992 ; SAPARBAEV et al. 1996 ; PAQUES and HABER 1997 ; SUGAWARA et al. 1997 ) as has been demonstrated for mitotic recombination. Alternatively, the recently described Rad1p/Rad10p independent pathway for removing nonhomologous tails may also be acting (COLAIACOVO et al. 1999 ). A mechanism not requiring 3' end degradation that also leads to sequence loss is cleavage of single-strand loops (Fig 6F) contained within heteroduplex DNA, as envisaged by FOGEL et al. 1984 to explain copy number changes at the CUP1 locus.

In conclusion, changes in minisatellite size during meiosis have been characterized at the sequence level. On the basis of our results we propose that degradation of the "recipient" DNA molecule occurs during gene conversion events and is not readily detected by the systems normally employed to investigate the mechanisms of recombination. In addition, we propose that the heteroduplex formed between newly synthesized and the recipient DNA can form by single strands of DNA annealing in regions of homology. Another interesting finding of this work is that there are recombination events involving the minisatellite inserts that do not result in any change to a novel size. Such events are not easily detected in mammalian systems, suggesting that mammalian minisatellites may be even more recombinationally active than previously supposed. Finally, these results demonstrate a clear association between minisatellite size changes and other recombination events, such as crossovers and gene conversions, in the flanking DNA region.

	ACKNOWLEDGMENTS

We are grateful to Dr. R. Kolodner, Dr. G. Carignani, Dr. M. Liskay, Dr. P. Watt, Dr. C. Falco, and Dr. N. Hunter for providing plasmids and to Dr. Janice Lamb for synthesizing oligonucleotides. We thank Professor John Clegg, Dr. John Armour, Dr. Mark Hirst, and members of the R.H.B. lab for helpful discussions. We also thank Dr. Fiona Pryde and Dr. Scott Chambers for comments on the manuscript. This work was supported by grants to R.H.B. and E.J.L. from the Wellcome Trust and a Medical Research Council Studentship to A.J.R.B.

Manuscript received November 13, 1998; Accepted for publication May 1, 2000.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

APPELGREN, H., H. CEDERBERG, and U. RANNUG, 1997  Mutations at the human minisatellite MS32 integrated in yeast occur with high frequency in meiosis and involve complex recombination events. Mol. Gen. Genet. 256:7-17[Medline].

APPELGREN, H., H. CEDERBERG, and U. RANNUG, 1999  Meiotic interallelic conversion at the human minisatellite MS32 in yeast triggers recombination in several chromatids. Gene 239:29-38[Medline].

ARMOUR, J. A., P. C. HARRIS, and A. J. JEFFREYS, 1993  Allelic diversity at minisatellite MS205 (D16S309): evidence for polarized variability. Hum. Mol. Genet. 2:1137-1145[Abstract/Free Full Text].

BORTS, R. H. and J. E. HABER, 1987  Meiotic recombination in yeast: alteration by multiple heterozygosities. Science 237:1459-1463[Abstract/Free Full Text].

BORTS, R. H. and J. E. HABER, 1989  Length and distribution of meiotic gene conversion tracts and crossovers in Saccharomyces cerevisiae.. Genetics 123:69-80[Abstract/Free Full Text].

BORTS, R. H., M. LICHTEN, M. HEARN, L. S. DAVIDOW, and J. E. HABER, 1984  Physical monitoring of meiotic recombination in Saccharomyces cerevisiae.. Cold Spring Harbor Symp. Quant. Biol. 49:67-76[Abstract/Free Full Text].

BORTS, R. H., M. LICHTEN, and J. E. HABER, 1986  Analysis of meiosis-defective mutations in yeast by physical monitoring of recombination. Genetics 113:551-567[Abstract/Free Full Text].

BORTS, R. H., W.-Y. LEUNG, K. KRAMER, B. KRAMER, and M. S. WILLIAMSON et al., 1990  Mismatch repair-induced meiotic recombination requires the PMS1 gene product. Genetics 124:573-584[Abstract].

BORTS, R. H., S. R. CHAMBERS, and M. F. F. ABDULLAH, 2000  The many faces of mismatch repair in meiosis. Mutat. Res. 451:129-150[Medline].

BUARD, J. and G. VERGNAUD, 1994  Complex recombination events at the hypermutable minisatellite CEB1 (D2S90). EMBO J. 13:3203-3210[Medline].

CAO, L., E. ALANI, and N. KLECKNER, 1990  A pathway for generation and processing of double-strand breaks during meiotic recombination in S. cerevisiae.. Cell 61:1089-1101[Medline].

CHAMBERS, S. R., N. HUNTER, E. J. LOUIS, and R. H. BORTS, 1996  The mismatch repair system reduces meiotic homeologous recombination and stimulates recombination-dependent chromosome loss. Mol. Cell. Biol. 16:6110-6120[Abstract].

CHEN, W. and S. JINKS-ROBERTSON, 1999  The role of the mismatch repair machinery in regulating mitotic and meiotic recombination between diverged sequences in yeast. Genetics 151:1299-1313[Abstract/Free Full Text].

COLAIACOVO, M. P., F. PAQUES, and J. E. HABER, 1999  Removal of one nonhomologous DNA end during gene conversion by a RAD1- and MSH2-independent pathway. Genetics 151:1409-1423[Abstract/Free Full Text].

COLLINS, I. and C. S. NEWLON, 1994  Meiosis-specific formation of joint DNA molecules containing sequences from homologous chromosomes. Cell 76:65-75[Medline].

CROUSE, G. F., 1998 Mismatch repair systems in Saccharomyces cerevisiae, pp. 411–448 in DNA Damage and Repair: Biochemistry, Genetics and Cell Biology, edited by M. F. HEOKSTRA and J. A. NICKOLOFF. Humana Press, Totowa, NY.

DEBRAUWERE, H., J. BUARD, J. TESSIER, D. AUBERT, and G. VERGNAUD et al., 1999  Meiotic instability of human minisatellite CEB1 in yeast requires DNA double-strand breaks. Nat. Genet. 23:367-371[Medline].

DOUTRIAUX, M.-P., R. WAGNER, and M. RADMAN, 1986  Mismatch-stimulated killing. Proc. Natl. Acad. Sci. USA 83:2576-2578[Abstract/Free Full Text].

FAN, H. Y., K. K. CHENG, and H. L. KLEIN, 1996  Mutations in the RNA polymerase II transcription machinery suppress the hyperrecombination mutant hpr1 of Saccharomyces cerevisiae.. Genetics 142:749-759[Abstract].

FISHMAN-LOBELL, J. and J. E. HABER, 1992  Removal of non-homologous ends in double-strand break recombination: the role of the yeast ultraviolet repair gene RAD1.. Science 258:480-448[Abstract/Free Full Text].

FOGEL, S., J. W. WELCH, and E. J. LOUIS, 1984  Meiotic gene conversion mediates gene amplification in yeast. Cold Spring Harbor Symp. Quant. Biol. 49:55-65[Abstract/Free Full Text].

FOUCAULT, F., J. BUARD, F. PRAZ, C. JAULIN, and D. STOPPA-LYONNET et al., 1996  Stability of microsatellites and minisatellites in Bloom syndrome, a human syndrome of genetic instability. Mutat. Res. 362:227-236[Medline].

GALLI, A. and R. H. SCHIESTL, 1995  On the mechanism of UV and gamma-ray-induced intrachromosomal recombination in yeast cells synchronized in different stages of the cell cycle. Mol. Gen. Genet. 248:301-310[Medline].

GEBHART, E., R. BAUER, U. RAUB, M. SCHINZEL, and K. W. RUPRECHT et al., 1988  Spontaneous and induced chromosomal instability in Werner syndrome. Hum. Genet. 80:135-139[Medline].

GEITZ, D., A. ST. JEAN, R. A. WOODS, and R. H. SCHIESTL, 1992  Improved method for high efficiency transformation of intact yeast cells. Nucleic Acids Res. 20:1425[Free Full Text].

GILL, P., A. J. JEFFREYS, and D. J. WERRETT, 1985  Forensic application of DNA `fingerprints'. Nature 318:577-579[Medline].

GOLDMAN, A. S. and M. LICHTEN, 1996  The efficiency of meiotic recombination between dispersed sequences in Saccharomyces cerevisiae depends upon their chromosomal location. Genetics 144:43-55[Abstract].

GRODEN, J. and J. GERMAN, 1992  Bloom's syndrome. XVIII. Hypermutability at a tandem-repeat locus. Hum. Genet. 90:360-367[Medline].

HASTINGS, P. J., 1988  Recombination in the eukaryotic nucleus. Bioessays 9:61-64[Medline].

HOROWITZ, H. and J. E. HABER, 1984  Subtelomeric regions of yeast chromosomes contain a 36 base-pair tandemly repeated sequence. Nucleic Acid Res. 12:7105-7121[Abstract/